|
|
|
 |
Xavier Goaoc
LORIA (INRIA Lorraine, France)
Friday, Oct. 7, 11:00am
LC 102, Brooklyn Campus, Polytechnic University
Abstract
Visibility problems arise in various application fields in computer
science, in particular in image synthesis where computing whether two
points or two surfaces are mutually visible is a common task.
Occlusions among opaque 3D objects are usually very complex. Thus,
the existing methods to solve 3D visibility problems are commonly
approximate and constitute, efficiency-wise, the bottleneck of many
applications, for example in global illumination algorithms such as
radiosity.
A so-called ``global approach'' to these problems emerged in the last
decade. It amounts to pre-computing most of the visibility
information and to organizing it in a data structure that can then be
efficiently queried to solve different problems. This approach raises
several issues. We need (i) a meaningful estimation of the amount of
visibility information contained in a scene, (ii) to design efficient
algorithms to compute this information and (iii) to make sure that
these algorithms handle geometric input in arbitrary degenerate
position in a correct way -- a task which is always difficult in
geometric computing.
I will present recent progress on these questions obtained by
studying the geometry and combinatorics of particular sets of lines
in space.
For further information please contact Hervé Brönnimann (hbr at poly.edu)